A central air conditioning billing method

Abstract

The application discloses a central air conditioner billing method, and offline establishment of first and second neural networks respectively models cooling characteristics of a central air conditioner terminal unit and different orientation room cooling characteristics, and training data samples of the two networks are collected when working conditions are approximately stable; online prediction of terminal unit cooling capacity and room cooling consumption is carried out based on the two trained network models, and the cooling consumption is proportionally corrected based on differences in cooling consumption characteristics of different orientation rooms under the same working condition, so that central air conditioner cost allocation is realized. The application can dynamically reflect cooling capacity changes of the central air conditioner terminal unit, and through model generalization calculation, a correction coefficient of different rooms excluding the influence of additional heat load is obtained, so that various influencing factors such as sunlight and adjacent room heat transfer are effectively compensated. The application can accurately measure actual cooling capacity of the central air conditioner terminal unit under different working conditions, and realizes fair billing according to room internal heat load cooling consumption, which is helpful to the energy saving and utilization of air conditioners.

Description

Technical Field

[0001] This invention relates to the field of central air conditioning metering and billing, and more specifically to a central air conditioning billing method. Background Technology

[0002] A central air conditioning system consists of a cold / heat source system and a cold / heat transfer and air conditioning system. In summer, the cooling system provides the necessary cooling capacity to the air conditioning system to offset the heat load of the indoor environment; in winter, the heating system provides the same cooling capacity. Central air conditioning systems supply cooling to each terminal unit via refrigerant pipes, where the refrigerant can be water, air, or a refrigerant. Taking a water-cooled central air conditioning system as an example, it uses a single main unit connected to multiple fan coil units via chilled water pipes to deliver cooling capacity to different rooms to achieve indoor air conditioning.

[0003] The most prominent feature of central air conditioning is that it provides a comfortable working and living environment. With the rapid development of air conditioning demand in large public buildings in China, more and more office buildings, shopping malls, serviced apartments and other buildings are beginning to install central air conditioning systems.

[0004] Currently, many places still use the traditional and simple area-based billing method for central air conditioning. This method is simple and convenient, but most people do not consider whether the air conditioning is energy-efficient, resulting in the phenomenon of air conditioning being turned on even when no one is using it, leading to energy waste.

[0005] If energy metering is adopted, charging only for what is used can raise people's awareness of energy conservation. Reasonable billing methods can change consumers' energy consumption habits; therefore, choosing and adopting appropriate central air conditioning billing methods is of great significance for energy conservation.

[0006] The initial billing method for central air conditioning was based on area allocation, which resulted in wasted electricity. To address this, in recent years, a new individual billing system based on central air conditioning has been proposed. This system aims to achieve reasonable billing by allocating costs to each user in a given room, thereby reducing significant electricity waste and high building energy consumption.

[0007] Among the new technologies for cooling capacity metering and allocation billing in central air conditioning systems, chilled water metering methods have emerged. These methods are further divided into two types. One method involves installing a water meter at the outlet of the fan coil unit to measure the chilled water flow rate within the unit. This method simply solves the problem of inconsistent usage, but it doesn't consider the inlet and outlet temperatures of the chilled water. The other method, conversely, assumes a constant chilled water flow rate and only measures the temperature difference between the inlet and outlet, billing as long as the air conditioning is on. These methods, including the later improved energy metering method, all measure a few fixed parameters at the terminal, failing to reflect the impact of overall changes in the central air conditioning system's operating conditions on the terminal cooling capacity supply, thus making it difficult to accurately reflect the user's true cooling capacity consumption.

[0008] Taking water-cooled central air conditioning as an example, another issue that needs to be considered in traditional cooling capacity metering is that the temperature difference between the supply and return water is much smaller than that for heating. Therefore, this requires higher measurement accuracy of the temperature sensor. Undoubtedly, using high-precision sensors at each terminal will greatly increase the user's cost; while using ordinary sensors generally suffers from large sampling fluctuations and inaccurate measurement.

[0009] High-cost energy meters face difficulties in widespread adoption. Currently, according to surveys, most newly built central air conditioning cooling capacity allocation and billing systems employ indirect or equivalent billing methods, with time-based billing being the most common. Time-based metering calculates the equivalent or cumulative cooling capacity under rated test conditions, as illustrated by Chinese patent CN 100504338C, which calculates the cumulative operating time of each fan speed setting at the terminal unit. Time-based billing assumes a direct proportionality between cooling capacity and the inlet and outlet water temperature difference of the fan coil unit, as well as the water volume. It considers fan speed as a factor influencing temperature difference, assuming that higher fan speeds result in higher air volumetric flow rates and consequently, higher cooling capacity equivalents. For systems with high, medium, and low fan speed settings... H V M V L For a constant fan coil unit, the equivalent cooling capacity is:

[0010] Q = K H t H +K M t M +K L t L ,

[0011] Among them, t H t M t L K represents the opening time (s) of the two-way valve under high, medium, and low wind speeds. H K M K LThis represents the proportionality coefficient (kJ / s) for high, medium, and low wind speeds. The two-way valve itself is a switching component. By detecting the opening and closing of each two-way valve, the cumulative opening time of the two-way valve within a certain period, as well as the different wind speed settings, can be obtained. Therefore, in the time-based metering method, the cooling capacity delivered by each fan coil unit of a water-cooled central air conditioning system is calculated based on the cumulative time of opening and closing the two-way valve at different wind speeds. The key to this method lies in how to obtain K. H K M K L The coefficient? The current method is to use estimated empirical or theoretical values, or to use the coefficient value calculated by the fan coil unit manufacturer based on rated conditions such as dry bulb temperature of 27°C, wet bulb temperature of 19.5°C, and chilled water inlet temperature of 7°C, and the cooling capacity at various fan speeds.

[0012] This method of calculating cooling capacity under dynamic operating conditions using a fixed coefficient is obviously only suitable for estimation and cannot be used for accurate measurement.

[0013] Central air conditioning system cooling capacity metering is the foundation of cooling capacity billing, but many factors need to be considered from metering to billing. For general buildings, cooling capacity costs cannot be simply calculated by multiplying the cooling capacity usage by the unit price of cooling capacity; the impact of different locations of air-conditioned rooms and heat transfer between units must also be considered.

[0014] Because room heat load is related to environmental parameters, solar radiation, and the heat transfer characteristics of the building envelope, the heat load of rooms with different orientations (e.g., south-facing, north-facing) and different heights (e.g., top floor, middle floor, ground floor) can vary significantly. Therefore, centralized cooling systems should adjust the cooling capacity or unit price for rooms based on their location and orientation. Furthermore, heat transfer between units is also a crucial factor. When adjacent rooms have low occupancy or are vacant, the cooling load of the air-conditioned room will increase. If charging is based solely on cooling capacity, the cost for rooms of the same area will differ significantly, which is unfair.

[0015] However, due to the combined effects of various factors such as sunlight, structure, and neighboring rooms, it is difficult to independently distinguish the degree of influence of each factor. Therefore, existing billing methods often rely on empirical coefficients. For example, in his dissertation "Research on Issues Related to Metered Heating," Tian Yuchen of Tianjin University, combining reference standards, used statistical methods to calculate correction coefficients for each household under different building structures, room types, and orientations. He also provided examples of correction coefficients for rooms in different locations in non-energy-efficient residences, giving correction coefficients ranging from 0.55 to 1.00 for nine types of rooms. This example-based method can only provide a rough guide and cannot accurately quantify the cooling loss relationship of rooms in different locations within a specific building.

[0016] Therefore, the industry urgently needs a system that can accurately measure the equivalent cooling capacity of central air conditioning terminals under actual operating conditions, and correct for cooling consumption in rooms of different locations. This system would allow for a reasonable allocation and billing method based on cooling consumption and with quantitative corrections for environmental factors, achieving the goal of paying more for more use and less for less use. Ultimately, this would raise energy conservation awareness among users and achieve the goals of energy conservation and environmental protection. Summary of the Invention

[0017] In view of this, the purpose of the present invention is to provide a method for accurately measuring the cooling capacity of the central air conditioning terminal unit in a central air conditioning system and making quantitative corrections for the cooling rooms before billing, so as to achieve reasonable and effective billing based on the actual cooling capacity.

[0018] The technical solution of the present invention is to provide a central air conditioning billing method, comprising the following steps:

[0019] S1. Initialization: Classify the central air conditioning terminal units and establish a first neural network for each category as the central air conditioning terminal unit metering mapping model. Classify the rooms cooled by the central air conditioning system by location and establish a second neural network for each location as the room cooling model.

[0020] The first neural network takes the operating parameters of the central air conditioning system as input and the equivalent cooling capacity per unit time of the terminal unit in this room as output.

[0021] The second neural network takes two scalars—the room temperature during cooling and the outdoor temperature—as input, along with a vector composed of the room's indoor temperature values ​​in six directions (front, back, left, right, up, and down) and outputs the equivalent cooling load of the room per unit time.

[0022] Based on the cooling capacity of the central air conditioning terminal unit, a reference air conditioning unit is selected as the comparison of cooling capacity. The sensible cooling capacity of the reference air conditioning unit under different operating conditions is obtained according to standard tests, and the parameters are recorded as operating characteristics.

[0023] S2. Collect data samples and train the first neural network.

[0024] The control terminal unit and the reference air conditioning unit cool independently in stages and alternately. In each stage, the room temperature and humidity are dynamically maintained at the preset target value. The cooling capacity per unit time calculated by the reference air conditioning unit under the same operating conditions is used as the equivalent cooling capacity of the terminal unit.

[0025] S3. For the second neural network, collect data samples and train it:

[0026] When the operating conditions are approximately stable, data samples of the second neural network under different input conditions are collected, wherein the sample output is predicted by the first neural network that has been trained, and training is performed based on the data samples;

[0027] S4. When using online applications, the cost of central air conditioning is allocated according to time periods:

[0028] First, based on the cooling loss model, calculate its correction coefficient under the current room conditions. In the formula, i is the current room number, where i = any integer from 1 to N, N is the total number of rooms, and p i p i0 To be respectively The mapping output of the second neural network when used as an input vector. The current operating status of room i. The room temperature and the outdoor temperature are in the same range as The room temperature values ​​in all six directions of the current room are the same, and the room temperature is taken as the room temperature.

[0029] The cost of room i in time period d is then calculated as follows:

[0030] in, The equivalent cooling capacity (PF) per unit time of the terminal unit in the current room, predicted based on the first neural network. i (t) The cumulative cooling capacity for this period, Q jd k is the cooling capacity supplied to room j during this period. j3 C is the coefficient corresponding to room j. d This represents the total unallocated cost of central air conditioning for this period.

[0031] In another embodiment of the present invention, a central air conditioning billing method is also provided, comprising the following steps:

[0032] S1. Initialization: Classify the central air conditioning terminal units and establish a first neural network for each category as the central air conditioning terminal unit metering mapping model. Classify the rooms cooled by the central air conditioning system by location and establish a second neural network for each location as the room cooling model.

[0033] The first neural network takes the operating parameters of the central air conditioning system as input and the equivalent cooling capacity per unit time of the terminal unit in this room as output.

[0034] The second neural network takes two scalars—the room temperature during cooling and the outdoor temperature—as input, along with a vector composed of the room's indoor temperature values ​​in six directions (front, back, left, right, up, and down) and outputs the equivalent cooling load of the room per unit time.

[0035] Based on the cooling capacity of the central air conditioning terminal unit, a reference air conditioning unit is selected as the comparison of cooling capacity. The sensible cooling capacity of the reference air conditioning unit under different operating conditions is obtained according to standard tests, and the parameters are recorded as operating characteristics.

[0036] S2. Collect data samples and train the first neural network: Control the terminal unit and the reference air conditioning unit to cool independently in stages and alternately. In each stage, the room temperature and humidity are dynamically maintained at the preset target value. The cooling capacity per unit time calculated by the reference air conditioning unit under the same working conditions is used as the equivalent cooling capacity of the terminal unit.

[0037] S3. For the second neural network, collect data samples and train it: When the operating conditions are approximately stable, collect data samples of the second neural network under different input conditions, wherein the sample output is predicted by the trained first neural network, and training is performed based on the data samples;

[0038] S4. When using online applications, the cost of central air conditioning is allocated according to time periods:

[0039] First, calculate the conversion factor based on the current room conditions. In the formula, i is the current room number, and s i s j The areas of rooms i and j are respectively, where i and j are any integers from 1 to N, and N is the total number of rooms. i p j Rooms i and j are respectively... When used as input vectors, each corresponds to the mapping output of the second neural network. The current operating status of room i.

[0040] And calculate the comfort factor. In the formula, p ic For room i When used as an input vector, it corresponds to the output of the second neural network. and The only difference is that the room temperature is a preset comfort temperature.

[0041] The cost of room i in time period d is then calculated as follows:

[0042] Where, k j1 k j2 C is the coefficient corresponding to room j. d This represents the total unallocated cost of central air conditioning for this period.

[0043] Preferably, the comfortable temperature can be set at 25 degrees Celsius when cooling and at 18 degrees Celsius when heating.

[0044] Preferably, the room number corresponding to the max() function can be obtained offline first.

[0045] Preferably, in the calculation of the correction coefficient for the d-th time period, the input vector of the second neural network... Each parameter is taken as the average value within that time period; The outdoor temperature can be taken as the room temperature.

[0046] Preferably, the cooling capacity Q for this period is... id It is obtained by discrete short-period accumulation.

[0047] Preferably, a solar radiation intensity parameter is added to the input of the second neural network.

[0048] Preferably, the central air conditioning is a water-cooled central air conditioning, the terminal unit is a terminal fan baffle, and the first neural network uses a vector composed of four scalars: chilled water supply flow rate of the central air conditioning unit, supply and return water temperature difference, current temperature and humidity of the room, and the opening status values ​​of all terminal fan baffles as input.

[0049] Preferably, the central air conditioning is an all-air central air conditioning, the terminal unit is a terminal fan baffle, and the first neural network uses a vector composed of four scalars: the air supply flow rate of the central air conditioning unit's air outlet, the supply and return air temperature difference, the current temperature and humidity of the room, and the opening status values ​​of all terminal fan baffles as input quantities.

[0050] Preferably, the central air conditioning system is a variable refrigerant flow central air conditioning system, the terminal unit is an indoor unit, and the first neural network uses a vector composed of four scalars: the power of the central air conditioning unit, the power of the indoor unit, the current temperature and humidity of the room, and the fan on / off status values ​​of all terminal units as input quantities.

[0051] Preferably, the data sample acquisition process in step S2 is as follows: taking the room where the central air conditioning terminal unit is located as the heat load, setting a target temperature based on the initial temperature of the room; controlling the operation of the central air conditioning terminal unit's fan driver to adjust the room temperature to the target temperature and obtaining the current humidity as the target humidity.

[0052] After the room has maintained the target temperature for a period of time, samples are collected by sequentially using the reference air conditioning unit, the central air conditioning terminal unit, and the reference air conditioning unit as individual cold sources for a certain period of time during three time periods. While each of the two cold sources is operating, the room temperature is maintained at the target temperature. The terminal unit operates in PWM mode, and the reference air conditioning unit, while operating, also maintains the humidity at the target humidity level through a condition adjustment unit.

[0053] Based on the operating characteristics and conditions of the reference air conditioning unit, the cooling power consumption is calculated and divided by the average PWM duty cycle of the central air conditioning terminal unit during its working period. The resulting value is used as the equivalent cooling capacity per unit time when the fan of the central air conditioning terminal unit is currently in operation (F).

[0054] Preferably, step S2 specifically includes:

[0055] S21. Turn off the central air conditioning terminal unit of this room and start timing. Control the reference air conditioning unit to maintain the room temperature at the target temperature from t=0 to t=T1 during the first time period. At the same time, control the operating condition adjustment unit to maintain the room humidity at the target humidity. Calculate the cooling power q(t) based on the operating characteristics and operating conditions, and accumulate the cooling capacity for the first time period.

[0056] S22. Turn off the reference air conditioning unit and start timing again. After setting the fan on / off state value F of the central air conditioning terminal unit, control its fan driver to work in PWM mode and maintain the room temperature at the target temperature from the second time period from t=0 to t=T2, and calculate its equivalent time. Where Δ(t) is the PWM value,

[0057] S23. Turn off the central air conditioning terminal unit again and restart the timing. Control the reference air conditioning unit to maintain the room temperature at the target temperature during the third time period from t=0 to t=T1. At the same time, control the operating condition adjustment unit to maintain the room humidity at the target humidity. Calculate the cooling capacity for the third time period again.

[0058] S24. Calculate the equivalent cooling capacity of the fan per unit time under the current operating conditions:

[0059] As a preferred option, when a room has multiple central air conditioning terminal units, the cooling load of the room during this period is... Where F i This is the collection of all windshields in the current room;

[0060] Preferably, the cost sharing of central air conditioning also includes a portion allocated by area, so the cost of room i in time period d is:

[0061]

[0062] Among them, C d1 and C d0 These are the operating costs and basic costs of the central air conditioning system.

[0063] Preferably, the data samples of the second neural network are automatically extracted during the operation of the central air conditioning system: the heat load conditions and cooling consumption parameters of the central air conditioning cooling room are periodically and continuously collected, and the cycles in which the input quantities of the second neural network change within a preset fluctuation threshold are searched out, and the data of the selected cycles are stored in the data sample set of the second neural network.

[0064] Preferably, the preset fluctuation threshold is ±5%.

[0065] Preferably, the preset fluctuation threshold includes a first fluctuation threshold and a second fluctuation threshold. When the change of the input quantity of the second neural network within a period exceeds the first fluctuation threshold but is less than the second fluctuation threshold, the sample value of the input quantity within this period is taken as its time average value within this period.

[0066] As a preferred option, outdoor temperature t w For example, the average outdoor temperature within period T, i.e., its time average, is... Preferably, the first fluctuation threshold and the second fluctuation threshold are ±2% and ±5%, respectively.

[0067] Preferably, the chilled water supply flow rate and the supply-return water temperature difference in the input of the first neural network are obtained by measuring points set at the adjacent ends of the chilled water supply and return water main pipes and the central air conditioning unit, respectively.

[0068] In the three time periods mentioned in step S2, the temperature difference of the temperature detection modules at multiple indoor measurement points is less than the temperature difference threshold through indoor air circulation. The temperature detection modules can be set at the same height and located on two vertical diagonals respectively. The temperature difference threshold can be a value between 0.1℃ and 0.5℃. The current temperature of the room is the average of the temperatures at multiple measurement points or the temperature value of the return air vent.

[0069] The humidity is sensed by a humidity detection module installed in the middle of the room's return air duct.

[0070] Preferably, the activation status value of the central air conditioning terminal unit can be taken as the normalized value of the three fan power corresponding to the low, medium and high fan speeds, respectively. For example, the highest power value is taken as 1, and the power of the other two speeds is proportionally converted.

[0071] Preferably, step S2 further includes:

[0072] Acquire room images, and extract orientation features and two mutually perpendicular diagonals through image processing. The orientation features include the distribution and length of the room's structural edges, as well as the direction and distance of the cold source air outlet, return air outlet, and temperature uniform module relative to each corner of the room.

[0073] Take a line connecting the location of the cold air outlet to the furthest point in the room it can reach or the position directly opposite it as one of the main diagonals, and take another line perpendicular to it as a secondary diagonal; then, plan a spiral trajectory with the main diagonal as the axis.

[0074] During the three time periods, the fan axis at the end of the temperature uniform module moves along the planned trajectory to deliver the cold air blown out by the central air conditioning terminal unit and / or reference air conditioning unit to various areas of the room until the temperature difference of multiple temperature detection modules in the room is less than the temperature difference threshold.

[0075] Preferably, the movement along the planned trajectory specifically includes: first, running at a constant speed during each time period of sample collection; then, according to the temperature distribution characteristics of multiple temperature measuring points in the room, changing the linear velocity of the terminal fan moving along the trajectory so that the linear velocity is inversely proportional to the temperature difference between the temperature at the corresponding trajectory point and the target temperature.

[0076] Preferably, in step S1, the operating characteristics of the reference air conditioning unit are obtained based on a room-type air enthalpy test device. The sensible cooling capacity of the reference air conditioning unit under different operating conditions is tested, and the operating parameters and the sensible cooling capacity are recorded as a table or curve of operating characteristics.

[0077] In step S2, when collecting the first neural network sample, the cooling capacity of the reference air conditioning unit under the current operating conditions is calculated by querying and interpolating the table or curve based on the current operating condition parameter values.

[0078] Preferably, all terminal air deflectors are further subdivided into categories based on their specific models and their horizontal and vertical distances from the inlet of the chilled water supply main pipe of the central air conditioning unit. A first neural network is established for each sub-category, and training samples are collected for each sub-category.

[0079] Preferably, the reference air conditioning unit is equipped with a dry-bulb and wet-bulb temperature detection module and an air supply volume detection module. The sensible cooling capacity under different operating conditions is obtained according to standard tests, and the parameters are recorded as a working characteristic table or curve. Based on the current dry-bulb and wet-bulb temperatures and air supply volume of the inlet and outlet air, the cooling capacity of the reference air conditioning unit in the current room is calculated by querying and interpolating the working characteristics.

[0080] Preferably, in step S2, when the central air conditioning terminal unit operates in PWM mode, its equivalent time is calculated by multiplying the normalized speed of the fan motor at different fan speeds by the integral of the duty cycle Δ(t) within the cycle. Where k(t) is the normalized rotational speed, which is 1 for the highest speed and the ratio of the rotational speed to the highest speed for other speeds.

[0081] Preferably, the first neural network is a backpropagation (BP) neural network, and its model is as follows:

[0082] The output of the j-th node in the hidden layer is

[0083] The output of the output layer is

[0084] Where x1~x4 are four scalars: chilled water supply flow rate of the central air conditioning unit, supply and return water temperature difference, current temperature and humidity of the room; x5~xn are the on / off status values ​​of all central air conditioning terminal units; the f() function is taken as the sigmoid function, w ij and v j These are the connection weights from the input layer to the hidden layer and the connection weights from the hidden layer to the output layer, θ. j θ and θ are the thresholds for the hidden and output layers, respectively, and n and k are the number of nodes in the input and hidden layers, respectively. Gradient descent is used for network training.

[0085] Preferably, the second neural network can also be a BP network, in which x1 to x2 are two scalars: the room temperature and the outdoor temperature, and x3 to x8 are the room temperatures in the six directions: front, back, left, right, up, down, and left and right. The output y(t) is the equivalent cooling capacity per unit time of the fan in the room.

[0086] Preferably, there are 3 to 6 temperature detection modules, and the height is approximately 2 meters.

[0087] Preferably, the temperature difference between the target temperature and the initial temperature is ≥5℃.

[0088] Preferably, the reference air conditioning unit is a heating and cooling air conditioner; if the ratio of the equivalent time dT to the duration T2 of the second time period is less than the duty cycle threshold Δs, when collecting the sample, the reference air conditioning unit is also controlled to operate in heating mode from τ=0 to τ=T3 within the second time period, and the heating equivalent is recorded. Accordingly, the equivalent cooling capacity per unit time of the fan baffle under the current operating conditions is calculated:

[0089]

[0090] Preferably, an electric heating module can also be set in the reference air conditioning unit, and the electric heating module can be controlled to heat from τ=0 to τ=T3 in time range, and the heat equivalent Q3=pr·T3 is recorded, where pr is the heating power of the electric heating module (kW i.e. kJ / s), and PF is calculated similarly.

[0091] Preferably, within the time ranges T1 and T2, the heating module can be turned on with known power and heat calculation can be performed.

[0092] Preferably, the rated cooling power of the reference air conditioning unit is 0.85 to 1.15 times the maximum cooling capacity of the terminal fan.

[0093] Preferably, the temperature equalization module includes a base, a vertical rotating shaft, a curved support arm, a horizontal rotating shaft, and a tiltable bracket with two sections of support arms movably connected by bolts. A telescopic support rod at an acute angle to the axis of the horizontal rotating shaft is connected between the outer ends of the two sections of support arms. A temperature equalization fan is supported at the end of the tiltable bracket, and a fan cover is provided on the back of the fan.

[0094] Preferably, image acquisition is performed through the image acquisition module in the sensing and detection unit, and the first neural network training samples are acquired through the control unit; the control unit includes an input module, a main processing module, an image processing module, a wind turbine processing module, a mapping module, and an output module, and the control unit is configured as follows:

[0095] The image processing module analyzes the room's orientation features based on the room images acquired by the image acquisition module and extracts two mutually perpendicular diagonals;

[0096] The main processing module responds to events and schedules other modules.

[0097] The fan processing module adjusts the duty cycle of the driver's PWM wave based on the average of multiple temperatures at different locations in the room.

[0098] The neural network is established in the mapping module. The input layer of the neural network receives the input from the main processing module, and the output of the output layer is transmitted to the iterative learning unit and the main processing module through the first connection array and the second connection array, respectively. When training the neural network offline, the iterative learning unit adjusts the connection weights of the neural network according to the actual value of the cooling equivalent of the windshield per unit time input by the main processing module and the neural network through the first connection array and the network output value, respectively. When measuring online, the first connection array is disconnected, the neural network predicts the cooling equivalent of the windshield per unit time and outputs it to the main processing module through the second connection array. The main processing module processes and analyzes the data and outputs it through the output module.

[0099] Compared with the prior art, the method of this invention has the following advantages: This invention uses the rooms where the central air conditioning terminal units are located in different locations as the heat load, and uses a pre-calibrated mobile reference air conditioning unit as a reference to measure and calibrate the cooling capacity equivalent of the central air conditioning terminal units; and uses the calibrated terminal units to measure the cooling power consumption per unit area of ​​rooms in different orientations under similar conditions of sunlight and influence from adjacent rooms, and obtains the correction coefficient for each room accordingly, thereby achieving the purpose of billing central air conditioning based on the actual cooling capacity consumption and comfort of the basic indoor heat load. This invention uses a vector composed of four scalars—key factors affecting the cooling capacity of central air conditioning terminal units, such as the chilled water supply flow rate of the main unit, the supply and return water temperature difference, the current temperature and humidity of the room, and the on / off status values ​​of all terminal units—as input. The output is the cooling capacity per unit time (equivalent to the cooling capacity) at the current fan speed in the room. A first neural network is established as the terminal unit metering mapping model. This model reflects the impact of actual changes in the central air conditioning system's operating conditions on the cooling capacity of the terminal units and dynamically reflects changes in the cooling capacity of the terminal units, overcoming the shortcomings of existing technologies that estimate the actual changing cooling capacity of the fan speed using fixed coefficients. Furthermore, the measurement points for the supply water temperature difference and flow rate are only set at the inlet and outlet of the main chilled water supply pipe of the central air conditioning unit, replacing the multi-point arrangement at each terminal. The detection of large flow rates relative to small flow rates at the terminals reduces relative error, and the significant reduction in measurement points allows for further reduction of metering error through the use of high-precision temperature difference detection. Furthermore, this invention categorizes fan terminals based on their model and the horizontal and vertical distances from the central air conditioning unit's chilled water supply main inlet. The model uses the on / off status values ​​of all terminal units as input, thus decoupling the mutual constraints between different fan terminals. In the shared billing process, the room cooling model, represented by a second neural network, uses temperature and radiation factors as input to compensate for differences in sunlight exposure in rooms with different orientations. The room temperatures of adjacent rooms in six directions are used as input to reflect the impact of neighboring rooms on cooling consumption. Finally, through correction coefficients, differences in the building envelope are also eliminated.

[0100] This invention can accurately measure the actual cooling capacity of central air conditioning terminal units under different operating conditions and realize billing based on the effective cooling capacity of indoor heat load, which helps to save on the use of air conditioning. Attached Figure Description

[0101] Figure 1 This is a flowchart of the method of the present invention;

[0102] Figure 2A , Figure 2B Schematic diagrams of two structures of a central air conditioning billing device and system using the method of the present invention; Figure 2C This is a schematic diagram of the control unit.

[0103] Figure 3AThis is a schematic diagram of a water-cooled central air conditioning system. Figure 3B Schematic diagram of a variable refrigerant flow central air conditioning system;

[0104] Figure 4A For reference, see the air conditioning unit cooling diagram. Figure 4B Schematic diagram of cooling supply for terminal fan units;

[0105] Figure 5A , Figure 5B This is a schematic diagram showing the distribution of the temperature detection module and the temperature uniformization process.

[0106] Figure 6 This is a schematic diagram of the temperature uniformity module;

[0107] Figure 7A This is a schematic diagram illustrating the principle of cold energy metering mapping. Figure 7B A schematic diagram illustrating the principle of room cooling conversion;

[0108] Figure 8A This is a schematic diagram of the mapping module. Figure 8B This is a schematic diagram of the first neural network structure;

[0109] Figure 9 A schematic diagram illustrating the principle of temperature regulation for the terminal unit.

[0110] In the diagram: 1000 Central air conditioning billing system, 100 Central air conditioning billing device, 200 Server, 300 Terminal unit / terminal fan, 400 Driver, 500 Chilled water pipes; 600 Reference air conditioning unit;

[0111] 120 sensing and detection unit, 130 operating condition adjustment unit, 140 user interface unit, 150 control unit;

[0112] 121 Temperature detection module, 122 Flow detection module, 123 Image acquisition module, 124 Electricity metering module;

[0113] 131 Temperature uniform module, 132 Vertical rotating shaft, 133 Bent-angle support arm, 134 Horizontal rotating shaft, 135 Telescopic support rod, 136 Tilting bracket, 137 Fan cover, 138 Temperature uniform fan, 139 Base.

[0114] 151 Input module, 152 Main processing module, 153 Image processing module, 154 Operating condition processing module, 155 Fan processing module, 156 Output module, 157 Storage module, 158 Mapping module;

[0115] 1521 Sample Extraction Department, 1522 Correction Coefficient Calculation Department, 1523 Billing Processing Department;

[0116] 1541 Temperature Uniformation Processing Unit, 1542 Humidity Control Unit;

[0117] 1581 First neural network, 1582 First connection matrix, 1583 Iterative learning unit, 1584 Second connection matrix;

[0118] 310 fan, 320 fan coil unit, 330 return air outlet; 340 indoor unit;

[0119] 610 outdoor unit module, 620 indoor unit module. Detailed Implementation

[0120] The preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings, but the present invention is not limited to these embodiments. The present invention covers any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of the present invention.

[0121] To provide the public with a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the invention, but those skilled in the art can fully understand the invention without these details.

[0122] The invention is described in more detail below by way of example with reference to the accompanying drawings. It should be noted that the drawings are in a simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0123] Example 1:

[0124] Central air conditioning is an air conditioning system that uses a single main unit or chiller to supply refrigerant to multiple cooling terminals in different rooms through ducts, water pipes, or refrigerant pipes to achieve indoor air conditioning. (Reference) Figure 3A As shown, water-cooled central air conditioning uses water as the refrigerant, while all-air central air conditioning is a VAV (Variable Air Volume) system. It controls and regulates the temperature of a specific air-conditioned area by changing the supply air volume rather than the supply air temperature, thus adapting to changes in the load of the air-conditioned area. See also... Figure 3B As shown, variable refrigerant flow (VRV) or VRF air conditioning systems control the refrigerant flow and achieve cooling or heating through direct evaporation or condensation of the refrigerant. Compared to the two heat exchange processes required by all-air and water-cooled central air conditioning systems, VRV or VRF central air conditioning systems only require one heat exchange, thus offering higher efficiency. However, their single-unit power is limited, making them suitable for small-scale centralized cooling / heating applications such as villas and partial floors of office buildings.

[0125] Combination Figure 1 , Figure 7B , Figure 7A As shown, the present invention provides a central air conditioning billing method, comprising the following steps:

[0126] S1. Initialization: Classify the central air conditioning terminal units and establish a first neural network for each category as the central air conditioning terminal unit metering mapping model. Classify the rooms cooled by the central air conditioning system by location and establish a second neural network for each location as the room cooling model.

[0127] The first neural network takes the operating parameters of the central air conditioning system as input and the equivalent cooling capacity per unit time of the terminal unit in this room as output.

[0128] The second neural network takes two scalars—the room temperature during cooling and the outdoor temperature—as input, along with a vector composed of the room's indoor temperature values ​​in six directions (front, back, left, right, up, and down) and outputs the equivalent cooling load per unit time for the room.

[0129] Based on the cooling capacity of the central air conditioning terminal unit, a reference air conditioning unit is selected as the comparison of cooling capacity. The sensible cooling capacity of the reference air conditioning unit under different operating conditions is obtained according to standard tests, and the parameters are recorded as operating characteristics.

[0130] S2. Collect data samples and train the first neural network.

[0131] The control terminal unit and the reference air conditioning unit cool independently in stages and alternately. In each stage, the room temperature and humidity are dynamically maintained at the preset target value. The cooling capacity per unit time calculated by the reference air conditioning unit under the same operating conditions is used as the equivalent cooling capacity of the terminal unit.

[0132] S3. For the second neural network, collect data samples and train it:

[0133] When the operating conditions are approximately stable, data samples of the second neural network under different input conditions are collected, wherein the sample output is predicted by the first neural network that has been trained, and training is performed based on the data samples;

[0134] S4. When using online applications, the cost of central air conditioning is allocated according to time periods:

[0135] First, based on the cooling loss model, calculate its correction coefficient under the current room conditions. In the formula, i is the current room number, where i = any integer from 1 to N, N is the total number of rooms, and p i p i0 To be respectively The mapping output of the second neural network when used as an input vector. The current operating status of room i. The room temperature and the outdoor temperature are in the same range as The room temperature values ​​in all six directions of the current room are the same, and the room temperature is taken as the room temperature.

[0136] The cost of room i in time period d is then calculated as follows:

[0137] in, The equivalent cooling capacity (PF) per unit time of the terminal unit in the current room, predicted based on the first neural network. i (t) The cumulative cooling capacity for this period, Q jd k is the cooling capacity supplied to room j during this period. j3 C is the coefficient corresponding to room j. d This represents the total unallocated cost of central air conditioning for this period.

[0138] Without loss of generality, this embodiment will first describe the cost-sharing method for a water-cooled central air conditioning system. In a water-cooled central air conditioning system, the terminal devices are terminal units, terminal fan units, or terminal windshields.

[0139] like Figure 2A As shown, the central air conditioning billing system 1000 using the method of the present invention includes a central air conditioning billing device 100, a server 200 for data communication and storage, and a driver 400 for driving the fans in the terminal units 300. The central air conditioning billing device 100 includes a control unit 150, and a sensing and detection unit 120, a user interface unit 140, and an operating condition adjustment unit 130 connected to the control unit 150. In this invention, the room where the water-cooled central air conditioning terminal unit 300 is located is used as the heat load. A reference air conditioning unit 600 is used as a reference for the varying cooling capacity of the room under different operating conditions. The sensing and detection unit 120 detects parameters of the operating conditions of the water-cooled central air conditioning system and the reference air conditioning unit 600, as well as the room's cooling conditions. The user interface unit 140 is used for inputting parameters and initiating operations, including displays during human-computer interaction, such as displaying the electricity costs allocated to each household.

[0140] A water-cooled central air conditioning system consists of one or more cold / heat source systems and multiple terminal air conditioning systems. The operation of a central air conditioning system is essentially a heat transfer process. Figure 3A , Figure 4B As shown, the cold / heat source is the main unit. Taking refrigeration as an example, the main unit uses a compressor for refrigeration. After passing through a heat exchanger, the circulating water is cooled into chilled water, which is then transported to each terminal air conditioning system, i.e., the fan coil unit 320 in the figure, through the chilled water pipe 500. After the chilled water is supplied to each user's room through the fan coil unit, the water temperature rises. It then circulates back to the heat exchanger and is evaporated by the compressor refrigerant to remove heat and cool down to chilled water, thus continuously removing heat from the room. At the same time, the compressor refrigerant is drawn into the compressor and compressed into high-pressure vapor before being discharged to the condenser. The outdoor fan or cooling water system removes the heat from the condenser, i.e., a secondary heat exchange occurs on the condenser, and the hot air that carries away the heat emitted by the condenser is discharged into the outdoor environment.

[0141] refer to Figure 4B As shown, the fan coil unit 320 is widely used in hotels, shopping malls, office buildings, hospitals, and other places. It is a working unit for heat exchange between indoor air and chilled water. Its working principle is that the fan draws indoor air or a mixture of indoor and outdoor air under the action of the fan 310. After flowing through the surface cooler, i.e., the curved pipe through which chilled water flows, it is cooled and sent into the room, thereby lowering the indoor temperature to meet people's comfort requirements. The cooled air blown in is heated by the heat from people, equipment, and the walls in the room, and then circulates back to the fan coil unit 320 through the return air vent 330 for heat exchange again.

[0142] In the heat transfer process of a central air conditioning system, the main unit delivers cooling capacity to each terminal unit. (Reference) Figure 7B As shown, in order to fairly charge for the cooling consumption of each room, two issues need to be addressed: how much cooling capacity does each room actually use?

[0143] Furthermore, how much difference is there in the level of comfort achieved by each room after consuming the same amount of cooling energy, due to factors such as sunlight, building envelope, and adjacent rooms? And how can this difference be corrected? At the same time, how can different users set different target temperatures and how can different temperatures be billed separately?

[0144] How much cooling capacity does each room consume through the terminal unit? This is the first question that must be answered when paying based on energy consumption.

[0145] Currently, the calculation of the cooling capacity of the terminal unit is based on monitoring the status of the three-speed switch of the fan. It is obtained by weighted summation of the working time of the high, medium and low fan speed settings. However, the weight value, i.e. the coefficient, of each setting can often only be obtained from the data calibrated by the manufacturer under rated operating conditions.

[0146] The limitations of this fixed-weighting method are obvious. First, due to factors such as installation conditions (distance from the main unit), the actual air volume of different fans may differ from the nominal value of each air volume setting. Second, and more importantly, the operating conditions of a water-cooled central air conditioning system, including each terminal unit, are dynamically changing. Using a fixed value to calculate an actually changing value is clearly unreasonable.

[0147] In each room where cooling is used, not only will there be a difference between the fan speed and the nominal value, but the total cooling output of the main unit will also vary. Furthermore, the distribution of this total cooling output by each terminal unit is not a simple linear proportional relationship, but rather there are mutual constraints between them, that is, there is a non-linear coupling relationship between each fan speed.

[0148] Therefore, this invention treats the water-cooled central air conditioning system, including each terminal unit, as a whole, and considers the distribution of cooling capacity among each terminal unit as a black box. Based on nonlinear modeling theory, it models the mapping relationship between the key operating conditions of the system and the cooling equivalent of the terminal unit.

[0149] To identify the model, a dataset for identification is needed. This includes obtaining the equivalent cooling capacity under different operating conditions. How is the equivalent cooling capacity obtained? Currently, the enthalpy difference method of the heat transfer medium is commonly used.

[0150] This method was first used in central air conditioning billing systems to construct heat meters to measure the amount of heat used for heating. The heat meter consists of a hot water flow meter, a pair of temperature sensors, and an integrator. Its working principle is that hot water supplied by a heat source flows into the heat exchange system at a relatively high temperature and flows out at a lower temperature. During this process, heat is provided to the user through heat exchange. The amount of heat received by the user within a certain time period can be calculated using the following equation:

[0151] E=∫K(Ts-Tr)dV,

[0152] Where E is the heat output of the heat exchange system, K is the correction coefficient for the specific gravity and specific heat of hot water, Ts and Tr are the supply and return water temperatures, respectively, and V is the flow rate of hot water through the heating system over a period of time.

[0153] The main errors in the enthalpy difference method stem from the measurement of the working fluid flow rate and the determination of its enthalpy value, especially with smaller flow rates. Similarly, there are methods that calculate cooling capacity by detecting the supply and return air on the air side. Compared to water-side metering, air-side metering reduces the accuracy requirements of temperature measurement equipment and instruments because the supply air temperature difference is significantly larger than the supply and return water temperature difference. However, both air-side and water-side metering currently primarily involve setting measurement points on the supply and return lines of the working fluid at the end. Due to the smaller flow rate at the end and the much greater fluctuations in temperature and flow parameters compared to the main unit, a trade-off exists between sensor accuracy and instrument cost.

[0154] Based on the above research, to improve the model's generalization ability and prediction accuracy, this invention uses a vector composed of four scalars—chilled water supply flow rate of the water-cooled central air conditioning unit, supply and return water temperature difference, current room temperature and humidity—and the on / off status values ​​of all terminal units as input quantities. The output quantity is the equivalent cooling capacity per unit time of the current fan speed in the room, i.e., the equivalent cooling capacity. A first neural network is established in the control unit as the terminal unit metering mapping model. Notably, the two key influencing factors, flow rate and temperature difference, require only one measurement point. More importantly, the measured flow rate is on the main pipeline, which is much larger than the terminal flow rate, effectively improving measurement accuracy.

[0155] This invention uses the room where the water-cooled central air conditioning terminal unit is located as the heat load, and uses a reference air conditioning unit as the reference object for the varying cooling capacity of the room under different operating conditions to obtain the equivalent value of cooling capacity in the data sample required for system identification.

[0156] Specifically, such as Figure 7A As shown, the reference air conditioning unit, i.e. the reference machine, is an integrated portable air conditioner. First, its sensible cooling capacity under different operating conditions is obtained according to standard tests, and the parameters are recorded as a working characteristic table or curve. A second mapping from the operating conditions of the reference machine to the cooling capacity is established, which provides a basis for calculating the sample cooling capacity in the room where it works at the fan end.

[0157] Then, under various input combinations, training samples for the established first neural network are acquired offline. Based on the characteristics of cooling capacity transfer from the main unit to the terminal units in a water-cooled central air conditioning system, and to address the impacts of system time delay and large inertia, [further details are needed]. Figure 4A , Figure 4B As shown, the control unit of this invention collects sample data of the end-unit metering mapping model in the following manner:

[0158] The control unit drives the fan corresponding to the water-cooled central air conditioning terminal unit through a driver;

[0159] Using the room where the water-cooled central air conditioning terminal unit is located as the heat load, a target temperature is set based on the room's initial temperature. The operating condition adjustment unit ensures that the temperature difference between multiple temperature detection modules located at different positions within the room is less than a temperature difference threshold. The control driver uses the central air conditioning terminal fan to adjust the room temperature to the target temperature and obtains the current humidity as the target humidity.

[0160] After the room temperature is maintained at the target temperature for a period of time, the drive is turned off and a timer is started. The reference air conditioning unit is controlled to maintain the room temperature at the target temperature from t=0 to t=T1 in the first time period. At the same time, the operating condition adjustment unit is controlled to maintain the room humidity at the target humidity. The cooling power q(t) is calculated based on the operating characteristics and conditions of the reference air conditioning unit, and the cooling capacity in the first time period is accumulated.

[0161] After turning off the reference air conditioning unit and timing again, and setting the fan speed on / off value F, the driver is controlled to operate in PWM mode to maintain the room temperature at the target temperature from t=0 to t=T2 in the second time period, and its equivalent time is calculated. Where Δ(t) is the PWM value,

[0162] The drive is turned off again and the timing is restarted. The reference air conditioning unit is controlled to maintain the room temperature at the target temperature during the third time period from t=0 to t=T1. At the same time, the operating condition adjustment unit is controlled to maintain the room humidity at the target humidity. The cooling capacity for the third time period is calculated again.

[0163] Calculate the equivalent cooling capacity per unit time of the fan baffle under the current operating conditions:

[0164] Preferably, the duration of the third time period can differ from that of the first time period, such as the difference being within 20%. Preferably, the activation state value F represents three speeds: low, medium, and high.

[0165] During continuous sample collection, data can be collected in the second time period simply by changing the operating conditions of the water-cooled central air conditioning system, while the first and third time periods can be re-collected every few samples. The first sample collection after changing the target temperature should be carried out sequentially in three time periods.

[0166] Combination Figure 7A As shown, this invention establishes a first neural network in the control unit as the first mapping from the operating state of the central air conditioning system's fan terminal unit to the cooling capacity / cooling capacity equivalent. To avoid errors caused by low flow rates at the terminal, the measurement parameters at the high flow rate point in the main duct are used as the input for the mapping. In the working room, the terminal unit and the reference air conditioning unit alternately cool, and the measurement of the terminal unit to be measured is based on equal cooling capacity. That is, the cooling capacity obtained by the reference air conditioning unit under the same operating conditions through the second mapping is used as the cooling capacity equivalent of the terminal unit. During this process, the heat load conditions of the two cold sources, i.e., the heat flux density flowing into the room under cooling conditions, are the same through sensing and control of the operating conditions to ensure the reliability of the equal measurement.

[0167] Without loss of generality, when the water-cooled central air conditioning terminal unit is operating stably online, the room humidity will remain relatively stable, mainly due to seasonal climate constraints, once the main unit settings are determined. Therefore, the current room humidity during central air conditioning cooling is used as the target humidity, and the room humidity is maintained at this target humidity when the reference air conditioning unit is cooling.

[0168] For room temperature, the target temperature needs to be set based on the initial temperature of the room under natural conditions without cooling. Preferably, the temperature difference between the target temperature and the initial temperature is ≥5℃, and the load rate of the terminal unit during sample collection is greater than the set value, such as enabling its power to reach 0.5 to 1 times the rated power.

[0169] When cooling is provided, the ratio of the cold air inlet area to the room surface area creates a temperature gradient within the room, which may cause a deviation in the heat load when the two cooling sources are operating. To reduce this heat load deviation, see [link to relevant documentation]. Figure 5A , Figure 5B and combined Figure 6 As shown, the present invention sets multiple temperature detection modules 121 at different locations in the room, and uses the temperature uniform module 131 in the operating condition adjustment unit to make the temperature difference between these temperature detection modules less than the temperature difference threshold.

[0170] Specifically, such as Figure 6 As shown, the operating condition adjustment unit includes a temperature equalization module 131, which comprises a base 139, a vertical rotation shaft 132, a curved support arm 133, a horizontal rotation shaft 134, and a tiltable bracket 136 with two sections of support arms movably connected by bolts. A telescopic support rod 135, forming an acute angle with the axis of the horizontal rotation shaft 134, is connected between the outer ends of the two support arms. A temperature equalization fan 138 is mounted at the end of the tiltable bracket 136. Preferably, the temperature equalization fan 138 has a fan cover 137 on its back. The sensing and detection unit includes an image acquisition module 123, which can be located at the bottom of the curved support arm 133, thereby acquiring a global image of the room through the rotation of the vertical rotation shaft 132.

[0171] See Figure 2C As shown, preferably, the control unit 150 includes an input module 151, a main processing module 152, an image processing module 153, an operating condition processing module 154, a fan processing module 155, a mapping module 158, and an output module 157. The operating condition processing module 154 further includes a temperature uniformity planning unit 1541 and a humidity adjustment unit 1542. The control unit is also configured to:

[0172] The main processing module responds to events and schedules other modules.

[0173] Based on the room images acquired by the image acquisition module, the image processing module 153 analyzes the room's orientation features and extracts two mutually perpendicular diagonals. The orientation features include the distribution and length of the room's structural edges, as well as the direction and distance of the cold source air outlet, return air outlet, and temperature equalization module 131 relative to the corners of the room.

[0174] Combination Figure 5A , Figure 5BAs shown, the temperature uniformity planning unit 1541 in the operating condition processing module 154 plans the operating trajectory of the temperature uniformity module 131 based on the above-mentioned orientation features, so that the axis of the temperature uniformity fan 138, i.e. the end it points to, moves in a spatial spiral to deliver the cold air blown out by the central air conditioning terminal unit and / or the reference air conditioning unit to each area of ​​the room until the temperature difference of multiple temperature detection modules in the room is less than the temperature difference threshold.

[0175] The trajectory planning can be based on the obtained room diagonals. A main diagonal is formed by connecting the location of the cold air outlet to the furthest point in the room it can reach, or to the opposite side. A secondary diagonal is formed by drawing a line perpendicular to this line. Then, a spiral trajectory is planned around the main diagonals. Figure 5A In the diagram, the black dot represents the air outlet located in the corner. A spiral curve, centered on the diagonal line of the dot, is planned to move around the inner wall of the cone as the target trajectory. The main diagonal line is the center line of the cone. Figure 5B In this design, the air outlet is located in the middle of one side wall. Using this side wall as the base of a cylinder, a spiral trajectory is planned around the inner wall of the cylinder, with the main diagonal being the center line of the cylinder. The planned trajectory must avoid the return air vent to prevent heat loss.

[0176] As a preferred approach, typical indoor orientation features can be summarized and classified, and the terminal trajectory curve of each orientation category can be planned based on geometric equations. The output angles of each joint in the temperature equalization module, including the vertical rotation axis, the horizontal rotation axis, and the telescopic support rod, can be analyzed based on inverse kinematics.

[0177] Alternatively, the joint angles corresponding to the trajectory can be stored as a data sequence through on-site teaching, and the joints can be controlled online in sequence according to this sequence.

[0178] Moving along a planned trajectory allows for rapid cooling of all areas of the room, reducing temperature gradients between them. The temperature uniformity planning unit controls the temperature uniformity module based on this planned trajectory. During each time period of sample collection, it first operates at a constant speed to achieve general cooling. To further reduce regional temperature differences, and as a preferred method, it then adjusts the linear velocity of the temperature uniformity fan along the planned trajectory based on the temperature characteristics of multiple temperature detection modules in the room. This linear velocity is inversely proportional to the temperature difference between the temperature at the corresponding trajectory point and the target temperature. The temperature at each trajectory point can be calculated through interpolation based on the temperature values ​​from multiple temperature measurement points. This speed planning improves the uniformity of room temperature across different spatial points, thereby ensuring consistent operating conditions and enhancing the generalization ability of the metrology model.

[0179] The temperature equalization module operates periodically. When the temperature difference between the highest and lowest temperatures among the multiple temperature detection modules is less than a temperature difference threshold, it stops operating and can collect data samples. When the temperature difference is detected to exceed the threshold, it restarts, thereby dynamically balancing the overall room temperature at the target temperature. Preferably, the temperature difference threshold is a value between 0.1℃ and 0.3℃.

[0180] During the second time period of sample collection, the fan processing module in the control unit adjusts the PWM wave duty cycle of the driver based on the average of multiple temperatures at different locations in the room, i.e., the overall room temperature. (Reference) Figure 9 As shown, based on the error value e(t) between the target temperature and the current overall room temperature, the fan processing module in the host unit calculates the PWM value connected to the fan driver based on the PID control law, and changes the fan speed by changing the drive power pulse width of the driver, so that the error value e(t) dynamically approaches 0.

[0181] The operating condition processing module is also equipped with a humidity control unit, which controls the operation of the humidity control module in the operating condition control unit based on the monitoring of humidity measurement points in the room, so that the humidity of the room is maintained at the target humidity.

[0182] Combination Figure 3A , Figure 5A , Figure 5B As shown, the sensing and detection unit has a humidity detection module in the middle of the room return air duct; a flow detection module 122 and a water temperature detection module 121 are installed at the inlet of the chilled water supply main pipe of the central air conditioning unit; a water temperature detection module is installed at the outlet of the return water main pipe connected to the unit; and multiple temperature detection modules are installed at different locations in the room, which can be installed at the same height and located on two vertical diagonals respectively.

[0183] Preferably, there are 3 to 6 temperature detection modules, set at a height of about 2 meters; the average value of the temperature values ​​from the multiple temperature detection modules can be used as the current temperature of the room.

[0184] During sample collection, the chilled water supply flow rate and supply-return water temperature difference of the central air conditioning unit can be kept constant by adjusting the operating power of the unit. Preferably, these two parameters can be taken as the average value during the sample collection period based on the ratio of cooling capacity to temperature difference and flow rate.

[0185] As a preferred option, the multiple temperature detection modules set up to control the heat load of the two cold sources are only used for sample collection. Therefore, one of the temperature detection modules, such as the module near the temperature measuring point of the return air vent, can be selected as the current room temperature in the first neural network input during online application, thereby simplifying the system structure and facilitating actual operation.

[0186] For each terminal unit of a water-cooled central air conditioning system, the fan speed opening value can be taken as the normalized value of the three fan power corresponding to the low, medium and high fan speed settings. For example, the highest power value is taken as 1, and the power of the other two settings is calculated proportionally.

[0187] The cooling system of a water-cooled central air conditioning system is a nonlinear hysteresis system. Therefore, changes in operating parameters require a period of time to reflect their impact. To address this, this invention sets sampling conditions when collecting training samples. Before sampling, the system is brought to a steady-state operating state. When the reference air conditioning unit and terminal unit are cooling, the room conditions are maintained for a period of time to eliminate the randomness of short-duration sampling. At the same time, by sampling the reference air conditioning unit once before and once after the terminal unit is cooling, the influence of slow fluctuations in operating conditions on the sampled data is eliminated, thereby improving the prediction accuracy of the network model.

[0188] The first neural network, trained using the collected sample set, is used to predict the equivalent cooling capacity per unit time of the current fan speed in the field environment. The predicted value is then output through the output module and can be used as the basis for billing each terminal of the water-cooled central air conditioning system.

[0189] like Figure 7A As shown, with the working room as the heat load, the equivalent cooling capacity provided by the terminal unit under the same operating conditions is obtained through a second mapping using a reference air conditioning unit. Therefore, the operating characteristics of the portable reference air conditioning unit must be obtained beforehand through high-precision calibration.

[0190] See Figure 4A As shown, the reference air conditioning unit 600 includes an outdoor unit module 610 and an indoor unit module 620. This reference air conditioning unit can be equipped with dry-bulb and wet-bulb temperature detection modules and airflow detection modules. A test apparatus based on the room air enthalpy method is established according to room air conditioning standards to test its sensible cooling capacity under different operating conditions, and the parameters are recorded as a working characteristic table or curve, serving as the working characteristics of the reference air conditioning unit. Then, when collecting samples, based on the current dry-bulb and wet-bulb temperatures and airflow of the inlet and outlet air, the cooling capacity of the reference air conditioning unit under the current operating conditions is calculated by querying and interpolating this working characteristic.

[0191] In the air enthalpy method, the formula for calculating sensible cooling is:

[0192] φ sc =q m ·c pa ·(t a1 -t a2 ) / V n ·(1+W n ),

[0193] Where, φ sc q represents the sensible cooling capacity (W). m The air supply volume (m³) at the measuring point 3 / s), V n The specific volume of moist air at the measuring point (m³) 3 / kg), W n t represents the air humidity at the measuring point. a1 and t a2 The return air and supply air temperatures (°C) and the specific heat capacity at constant pressure (c) are respectively. pa =1005 + 1846W n (J / (kg·K)).

[0194] In the reference air conditioning unit, dry-bulb and wet-bulb temperatures are detected by sensors placed in an insulated section at the supply and return air vents, respectively. The dry-bulb and wet-bulb sensors are used to detect supply air temperature, return air temperature, and humidity; these parameters can also be obtained using temperature and relative humidity sensors. During training sample collection, the reference air conditioning unit, under the command of the control unit, adjusts its operating frequency to maintain the room temperature at the target temperature.

[0195] Preferably, the operating characteristic data of the reference air conditioning unit is recorded in tabular form, and the sensible cooling capacity under the current operating condition is calculated based on the table lookup and multidimensional interpolation.

[0196] Preferably, the rated cooling power of the reference air conditioning unit is 0.85 to 1.15 times the maximum cooling capacity of the terminal fan.

[0197] Combination Figure 8A , Figure 8B As shown, the control unit establishes a first neural network 1581 in the mapping module 158 as a metering mapping model for the end unit. The input layer of this first neural network receives input from the main processing module 152, and the output of the output layer is transmitted to the iterative learning unit 1583 and the main processing module 152 through the first connection array 1582 and the second connection array 1584, respectively. When training the first neural network offline, the iterative learning unit 1583 adjusts the connection weights of the first neural network 1581 until the learning ends, based on the actual value of the cooling equivalent of the windshield per unit time and the network output value input by the main processing module 152 and the first neural network 1581 through the first connection array 1582, respectively. During online metering, the first connection array 1582 is disconnected, and the first neural network 1581 predicts the cooling equivalent of the windshield per unit time and outputs it to the main processing module 152 through the second connection array 1584. The main processing module 152 processes and analyzes the data and outputs it through the output module 156. Figure 2A , Figure 2CAs shown, the output module can transmit the metering results to the user interface unit 140 for display or store them in the server 200, where the server 200 can communicate with one or more systems applying the method of this invention through a cloud platform. In the main processing module 152, the sample extraction unit 1521 controls the collection and screening of training samples, the correction coefficient calculation unit 1522 calculates each coefficient for the apportionment billing, and the billing processing unit 1523 performs the apportionment processing of central air conditioning costs based on the operating conditions of each terminal unit in a distributed processing manner.

[0198] Preferably, the first neural network is a BP first neural network, and its model is as follows:

[0199] The output of the j-th node in the hidden layer is

[0200] The output of the output layer is

[0201] Where x1~x4 are four scalars: chilled water supply flow rate of the central air conditioning unit, supply and return water temperature difference, current temperature and humidity of the room; x5~xn are the on / off status values ​​of all terminal units; the f() function is taken as the sigmoid function, w ij and v j These are the connection weights from the input layer to the hidden layer and the connection weights from the hidden layer to the output layer, θ. j θ and θ are the thresholds for the hidden and output layers, respectively, and n and k are the number of nodes in the input and hidden layers, respectively. Gradient descent is used for network training.

[0202] As an alternative, other redundancy factors, such as the air supply temperature of the terminal unit, can be added to the network input.

[0203] Because the chilled water supply pipeline is shared, and the pipeline water pressure decreases progressively with the distribution of chilled water, the actual cooling capacity of the water-cooled central air conditioning fan coil unit terminal unit depends not only on its model and fan speed setting, but also on its distance from the main unit. Therefore, for terminal units of the same model, they can be further subdivided into smaller categories based on the horizontal and vertical distances of the terminal fan speed setting from the inlet of the chilled water supply pipeline of the central air conditioning main unit. A first neural network is established for each subcategory, and training samples are collected separately, thereby making the prediction and measurement of cooling capacity equivalent more accurate.

[0204] Preferably, for each category, one terminal unit is selected for sample acquisition and training. For the first neural network of each category's terminal unit, when collecting training samples, a room with slow and small external temperature changes is selected as the heat load, thus allowing for continuous sample collection and shortening the overall sample set sampling time. Therefore, rooms inside a building can be used as the collection environment. Preferably, for rooms located at the corners of a building, sample collection is conducted at night or during periods of daytime without direct sunlight.

[0205] Cooling capacity metering is the basis for billing central air conditioning systems based on cooling capacity, but in addition to metering, billing also needs to consider many other factors. For example... Figure 7B As shown, the cooling capacity required per unit area for cooling varies significantly depending on the location of the room within a building. For example, the ground floor lobby has less enclosing structure and supplies more cool air to the environment; while rooms on the edges and top receive significantly more solar radiation than rooms in the center, thus requiring more cooling capacity. Clearly, simply charging based on actual cooling capacity consumption is unfair to these rooms that consume more cooling due to environmental factors.

[0206] Therefore, it is essential to first analyze the cooling characteristics of rooms in different locations within a building and then calculate the corresponding cooling capacity compensation. This requires, first and foremost, studying how to describe the cooling characteristics of rooms in different locations—for example, whether a model can be established, and how to establish one. This model must reflect the main, quantifiable factors that cause variations in cooling consumption in different locations, and these factors must be convertible to adjust or compensate for different rooms. Then, even if a model can be established, how should it be modified? Extensive analysis has revealed that this modification should reflect the influence of environmental factors and exclude the current setpoint or actual temperature of each room from the modification itself; otherwise, it would revert to a billing method where regardless of cooling usage, the amount of cooling consumed remains the same.

[0207] For this reason, see Figure 7B As shown, this invention uses two scalar quantities—the current room temperature and the outdoor temperature—as well as a vector composed of the room's indoor temperature values ​​in six directions (front, back, left, right, up, and down) during cooling as input quantities. The equivalent cooling capacity per unit time of the room's fan speed is used as the output quantity. A second neural network is established in the control unit for each room, serving as a cooling consumption model for each room. Then, based on this model, the equivalent cooling capacity per unit time for each room under near-optimal operating conditions is predicted, and a correction coefficient is calculated accordingly. Finally, based on this correction coefficient, the actual cooling capacity predicted by the first neural network is corrected, and the deferred costs of the entire water-cooled central air conditioning system are allocated accordingly.

[0208] Specifically, refer to Figure 8A , Figure 8B The modeling process of the first neural network is described, and the second neural network can also use a BP network. Of its nine inputs, x1-x2 represent the current room temperature, outdoor temperature, and solar radiation intensity; x3-x8 represent the room temperature in six directions (front, back, left, right, top, bottom, and front / back). The output y(t) represents the equivalent cooling load per unit time for the room. To enable the prediction and calculation of this equivalent cooling load (i.e., the required cooling load) using the cooling equivalent of the first neural network, this invention maintains a dynamically stable room temperature during sample collection to achieve a balance between cooling supply and demand. Therefore, the output of the second neural network is the equivalent cooling load per unit time for the room's fan speed. When the terminal unit is cooling, room humidity is mainly affected by seasonal climate; however, preferably, the second neural network also includes indoor humidity as an input. When adjacent rooms are empty, the temperature in that direction, such as the outdoor temperature, is used as the indoor temperature of that room. Preferably, when the water-cooled central air conditioning system enters a stable operating state, the six-directional room temperature values ​​in the second neural network can also be replaced with the fan operation status values ​​of the six-directional room's terminal unit. As a preferred option, a solar radiation intensity parameter is added to the input of the second neural network.

[0209] When the operating conditions are approximately stable, data samples of the second neural network under different input conditions are collected. The output is predicted by the first neural network corresponding to the end unit. The second neural network is trained based on the data samples and is used to predict the equivalent cooling load per unit time of each room under the current operating conditions.

[0210] To achieve automatic data sample acquisition for the second neural network, this invention includes a sample extraction module in the main processing module of the control unit. This module periodically and continuously collects the heat load and cooling consumption parameters of the water-cooled central air conditioning room. Based on the input quantities of the second neural network, it searches for periods where the variation amplitude of these variables is within a preset fluctuation threshold, and stores the data of the selected periods into the data sample set of the second neural network. This automatic sample extraction allows for sample acquisition during the operation of the water-cooled central air conditioning system, eliminating the need for lengthy dedicated startup and debugging for sample acquisition, thus significantly reducing data acquisition costs.

[0211] Preferably, the preset fluctuation threshold is ±5%, and the values ​​of each input sample parameter are the arithmetic mean or median value within the period.

[0212] Preferably, the preset fluctuation threshold may further include two sets of threshold ranges: a first fluctuation threshold and a second fluctuation threshold. When the change in the input quantity of the second neural network within a period exceeds the first fluctuation threshold but is less than the second fluctuation threshold, the sample value of the input quantity within this period is taken as its time average within this period. (Using outdoor temperature t...)w For example, the average outdoor temperature within period T, i.e., its time average, is...

[0213] Preferably, the first fluctuation threshold and the second fluctuation threshold are ±2% and ±5%, respectively.

[0214] When building the second neural network, a separate model can be created for each room; alternatively, rooms with similar locational characteristics, such as rooms of the same type in the middle, can be represented by the same network. By sharing models, the sample size can be further reduced, saving computation and modeling time.

[0215] After completing the above two modeling steps offline, the cost of the water-cooled central air conditioning system is allocated in time periods during online application. Unlike existing methods, this invention does not use fixed, statistically based empirical values ​​as coefficients when correcting the cooling load of rooms in different locations, but instead performs real-time calculations based on the actual operating conditions of each room.

[0216] Specifically, the correction coefficient is first calculated based on the cooling loss model and the current room conditions. In the formula, i is the current room number, where i = any integer from 1 to N, N is the total number of rooms, and p i p i0 To be respectively The mapping output of the second neural network when used as an input vector. The current operating status of room i. The current room temperature and outdoor temperature are compared with The room temperature values ​​in all six directions of the current room are the same, and the room temperature is taken as the current room temperature.

[0217] The cost of room i in time period d is then calculated as follows:

[0218] in, The equivalent cooling capacity (PF) per unit time of the terminal unit in the current room, predicted based on the first neural network. i (t) The cumulative cooling capacity for this period, Q jd k is the cooling capacity supplied to room j during this period. j3 C is the coefficient corresponding to room j. d This represents the total unallocated cost of central air conditioning for this period.

[0219] The correction factor inevitably involves the cooling characteristics and cooling equivalent of other rooms, and these rooms will have different room temperatures at the same time due to differences in operating time, environment, and set temperature. Therefore, when calculating the correction factor for a particular room, should the cooling equivalent be obtained based on its specific operating conditions?

[0220] Through in-depth research, this invention utilizes the generalization characteristics of room cooling models to uniformly predict the cooling demand of the same room under conditions where there is no additional heat load from adjacent rooms. The predicted value is essentially the cooling demand of the room's basic heat load, and the ratio of the cooling demand corresponding to this basic heat load to the actual cooling demand is used as a correction coefficient k. i3 In the calculation formula for the cost allocation of central air conditioning, the actual cooling capacity of the room is multiplied by the aforementioned correction factor k. i3 This allows for the allocation and billing of cooling requirements corresponding to the basic heat load within each room, ensuring fairness in the billing process. It can be understood that, according to the aforementioned cost allocation formula, rooms near the building edges, on the top or bottom floors, or those with high external heat loads introduced by the building envelope, have a corresponding correction factor k. i3 The coefficient will be smaller than that of the central room of the building, so that the billing compensation can be obtained through that coefficient.

[0221] Meanwhile, if the set temperature for the same room is lower at different times, the actual cooling load will inevitably be higher, thus spreading higher costs. Therefore, this device also incorporates comfort-based billing, which helps guide rational consumption of cooling and achieves energy-saving effects.

[0222] Preferably, in the calculation of the correction coefficient for the d-th time period, the input vector of the second neural network... The values ​​of each parameter are taken as the average values ​​within this time period; while the cooling capacity Q for this time period is... id It can be obtained by short-period accumulation in a discrete periodic system. Preferably, The outdoor temperature can be taken as the current room temperature. The total unallocated costs can be calculated based on the system's electricity, water, and other consumption over time.

[0223] When a room has multiple terminal units, the cooling load of the room in this time period Where F i This is the set of all windshields in the current room.

[0224] It is understandable that, compared to water-cooled central air conditioning, all-air central air conditioning uses air instead of cooling water as the refrigerant. Therefore, the billing device scheme for water-cooled central air conditioning in this embodiment is also applicable to all-air central air conditioning. Due to the detailed differences in the refrigeration process, the first neural network established in the control unit of the billing device is modified to use a vector composed of four scalars: the supply airflow rate of the all-air central air conditioning unit's air outlet, the supply and return air temperature difference, the current temperature and humidity of the room, and the on / off status values ​​of all terminal fan dampers as input quantities.

[0225] Example 2:

[0226] During the process of collecting training samples for the first neural network, if the initial target temperature is set too high or too low, or if the power of the terminal unit does not match the heat load of the room, the terminal fan will be limited to low power or high power operation when collecting samples.

[0227] Therefore, in order to ensure that the sample covers different load rates of the terminal fans from low to high, in this embodiment, when the control unit controls the driver to operate in PWM mode, it switches the terminal unit state to operate at different speeds during a sample acquisition period. The equivalent time is calculated by multiplying the normalized speed of the fan motor at different speeds by the integral of the duty cycle Δ(t) within the period. Where k(t) is the normalized rotational speed, which is 1 for the highest speed and the ratio of the rotational speed to the highest speed for other speeds.

[0228] For the same reason, this embodiment can also use the following method to obtain training samples. The reference air conditioning unit is a heating and cooling air conditioner; if the ratio of the equivalent time dT to the duration T2 of the second time period is less than the duty cycle threshold Δs, when collecting the samples, within the second time period, the reference air conditioning unit is also controlled to operate in heating mode from τ=0 to τ=T3, and the heating equivalent is recorded. Accordingly, the equivalent cooling capacity per unit time of the fan baffle under the current operating conditions is calculated:

[0229]

[0230] Preferably, an electric heating module can also be set in the reference air conditioning unit, and the electric heating module is controlled to heat from τ=0 to τ=T3 in time range, and the heat equivalent Q3=pr·T3 is recorded, where pr is the heating power of the electric heating module (kW or kJ / s), and PF is calculated similarly.

[0231] Preferably, within the time ranges T1 and T2, the heating module can be turned on with a known power and heat calculation can be performed, thereby expanding the operating condition coverage of the reference air conditioning unit and the water-cooled central air conditioning terminal unit.

[0232] The reference air conditioning unit's operating characteristics represent a mapping between its operating conditions and sensible cooling capacity. In this embodiment, this mapping can also be represented by a third neural network. The inputs to this network can be selected as the dry and wet bulb temperatures of the inlet and outlet air and the air volume; or preferably, electrical power, outdoor condenser temperature, indoor evaporator temperature and humidity; or preferably, compressor operating frequency, supply fan motor power, supply and return air temperature and humidity.

[0233] Considering the costs of central air conditioning equipment and installation, as well as the presence of some public rooms in the building, some existing charges are composed of a combination of fixed costs and operating costs. The costs incurred by the air conditioning system mainly include the following: a) Electricity costs for the air conditioning system (including electricity costs for chillers, cooling towers, water pumps, etc.); b) Water consumption for supplementary use of the air conditioning system; c) Labor and management costs for the operation of the air conditioning system; d) Depreciation costs for the air conditioning system equipment; e) Maintenance and repair costs for the air conditioning system; f) Other surcharges and property management fees. Of these, the first three items constitute the operating costs of the central air conditioning system, while the latter three remain essentially fixed during the operation of the air conditioning system and are referred to as basic costs.

[0234] Unlike Implementation Example 1, this embodiment uses a combined charging method, dividing the total deferred cost of the water-cooled central air conditioning system into two parts: in addition to the allocation based on actual cooling consumption, it also includes an allocation based on area. Therefore, the cost for room i in time period d is:

[0235]

[0236] Among them, C d1 and C d0 These are the operating costs and basic costs of the central air conditioning system.

[0237] Example 3:

[0238] This embodiment provides another method for billing central air conditioning, and exemplifies a variable refrigerant flow multi-split air conditioner as a central air conditioning system.

[0239] Variable refrigerant flow (VRV) central air conditioning, also known as multi-split central air conditioning system, consists of one outdoor unit / unit connected to several direct evaporation indoor units of different or the same type and capacity, forming a single refrigeration / heating cycle air conditioning system.

[0240] See Figure 3B As shown, VRF originally means direct expansion. There are no large-diameter air ducts or water pipes between the main unit and the indoor unit. Instead, a thin copper pipe is used to transport the low-temperature liquid refrigerant, which has been compressed and expanded by the outdoor unit, through a long-distance pipeline to the indoor unit, which is the terminal unit. In the indoor unit, the refrigerant enters the evaporator through the electronic expansion valve and quickly turns into a gas, while carrying away the heat at the same time.

[0241] In a VRF system, the refrigerant flow rate is variable, depending on the indoor load. When the indoor load is high, a larger amount of refrigerant is delivered to the indoor units to provide more cooling / heating; conversely, the load decreases. The compressor in the main unit regulates the refrigerant flow rate required for the entire system, while the refrigerant flow rate of individual indoor units is controlled by electronic expansion valves. Within the indoor units, fans also regulate the airflow to the room.

[0242] When distinguishing from Example 1, see [link to Example 1]. Figure 2B As shown, in a VRF system, the pipes used for refrigerant delivery are much smaller in diameter and have a smaller flow rate compared to water pipes or air ducts. Therefore, an electricity metering module 124 is connected to the power supply circuit of each indoor unit 340, which serves as the terminal unit.

[0243] Accordingly, when performing metering modeling on the cooling equivalent of the terminal unit, the first neural network uses a vector composed of four scalars: the power of the central air conditioning unit, the power of the indoor unit, the current temperature and humidity of the room, and the operating status values ​​of all terminal unit fans as input, and the cooling equivalent of the unit's cooling capacity per unit time as output.

[0244] This embodiment provides a central air conditioning billing method, which includes the following steps:

[0245] S1. Initialization: Classify the central air conditioning terminal units and establish a first neural network for each category as the central air conditioning terminal unit metering mapping model. Classify the rooms cooled by the central air conditioning system by location and establish a second neural network for each location as the room cooling model.

[0246] The first neural network takes the operating parameters of the central air conditioning system as input and the equivalent cooling capacity per unit time of the terminal unit in this room as output.

[0247] The second neural network takes two scalars—the room temperature during cooling and the outdoor temperature—as input, along with a vector composed of the room's indoor temperature values ​​in six directions (front, back, left, right, up, and down) and outputs the equivalent cooling load per unit time for the room.

[0248] Based on the cooling capacity of the central air conditioning terminal unit, a reference air conditioning unit is selected as the comparison of cooling capacity. The sensible cooling capacity of the reference air conditioning unit under different operating conditions is obtained according to standard tests, and the parameters are recorded as operating characteristics.

[0249] S2. Collect data samples and train the first neural network: Control the terminal unit and the reference air conditioning unit to cool independently in stages and alternately. In each stage, the room temperature and humidity are dynamically maintained at the preset target value. The cooling capacity per unit time calculated by the reference air conditioning unit under the same working conditions is used as the equivalent cooling capacity of the terminal unit.

[0250] S3. For the second neural network, collect data samples and train it: When the operating conditions are approximately stable, collect data samples of the second neural network under different input conditions, wherein the sample output is predicted by the trained first neural network, and training is performed based on the data samples;

[0251] S4. When using online applications, the cost of central air conditioning is allocated according to time periods:

[0252] First, calculate the conversion factor based on the current room conditions. In the formula, i is the current room number, and s i s j The areas of rooms i and j are respectively, where i and j are any integers from 1 to N, and N is the total number of rooms. i p j Rooms i and j are respectively... When used as input vectors, each corresponds to the mapping output of the second neural network. The current operating status of room i.

[0253] And calculate the comfort factor. In the formula, p ic For room i When used as an input vector, it corresponds to the output of the second neural network. and The only difference is that the room temperature is a preset comfort temperature.

[0254] The cost of room i in time period d is then calculated as follows:

[0255] Where, k j1 k j2 C is the coefficient corresponding to room j. d This represents the total unallocated cost of central air conditioning for this period.

[0256] In calculating the conversion factor k i1 When calculating the cooling load, the `max()` function is used to find the room with the highest cooling load per unit area. Since the input values ​​for obtaining the cooling load are the same (i.e., the operating conditions are the same), the denominator corresponds to the room with the highest temperature and cooling load per unit area, achieving the same level of comfort under comparable external heat loads; that is, the denominator corresponds to the room with the highest internal heat load. Therefore, the conversion factor k... i1 The meaning is the ratio of the internal heat load per unit area of ​​the current room i to the room with the largest internal heat load.

[0257] Since achieving a lower indoor temperature requires more cooling energy, this embodiment also introduces a comfort coefficient, which is a proportionality coefficient k representing the amount of cooling energy required to achieve the current indoor temperature relative to a preset comfort temperature, under the same operating conditions in the same room. i2 For example, this comfortable temperature could be set at 25 degrees Celsius when cooling and at 18 degrees Celsius when heating.

[0258] This embodiment is based on the conversion factor k. i1The current operating conditions of each room are scaled down to the room with the highest internal load, and the influence of heat load caused by different orientations or locations is cleverly avoided through proportional calculations; furthermore, based on the generalization of the cooling model, the comfort factor k is used... i2 The calculation incorporates rewards and penalties to compensate for the cooling consumption at different indoor temperatures. This ensures that the final billing can overcome the effects of room orientation and temperature settings. The model itself uses factors such as air temperature and the room temperatures of adjacent rooms in six directions as inputs, reflecting the differences in sunlight exposure in rooms with different orientations and the impact of adjacent rooms on cooling consumption. Furthermore, through normalization correction of the equivalent cooling consumption per unit area in different rooms under the same operating conditions, and proportional correction of the cooling consumption at different target room temperatures in the same room, differences in orientation, building envelope, and actual temperature are all compensated for.

[0259] To achieve fair billing based on cooling consumption, this invention first models the cooling capacity characteristics of terminal units and the cooling consumption characteristics of rooms in different orientations. Based on two nonlinear models, it predicts and measures the cooling capacity of terminal units and the cooling consumption of rooms under actual operating conditions. The predicted actual cooling consumption is then corrected based on the cooling consumption characteristics of rooms in different orientations to overcome differences in environmental influences. In the first modeling step, to avoid errors caused by detecting small airflow at the terminal units during online application, this invention uses the main chilled water supply pipe of the central air conditioning system as a measuring point, detecting its flow rate and the supply-return water temperature difference as inputs to the fan speed metering mapping model. Furthermore, through consistent control of heat load conditions, a reference air conditioning unit is used as the calculation reference for the cooling capacity supplied by the terminal units. The nonlinear mapping model is trained using a sample set, and the trained model is used to predict and calculate the cooling capacity of the central air conditioning terminal units during online application. In the second modeling step, through the design of the model parameter structure and correction coefficient formula, effective compensation is achieved for the influence of sunlight, building structure, and adjacent rooms. This invention can guarantee metering accuracy without increasing costs, and it decouples the mutual constraints between the various fan ends, enabling accurate measurement of dynamically changing cooling capacity. Furthermore, by introducing compensation based on actual operating conditions in the calculation of cooling capacity correction for rooms in different orientations, it ensures fair and reasonable billing.

[0260] It is understood that by interchangeding the cooling and heating operating conditions in this invention, this invention is also applicable to the billing of central air conditioning terminal units during the heating season.

[0261] The foregoing has described several embodiments of the present invention, but these embodiments are merely illustrative examples and do not limit the scope of the invention. These embodiments can be implemented in various other ways, and various omissions, substitutions, combinations, and modifications can be made without departing from the spirit of the invention. These embodiments or their variations are included within the scope or spirit of the invention, and are similarly included within the scope of the invention as described in the claims and its equivalents.

Claims

1. A central air conditioning billing method, comprising the following steps: S1, initialization, classifying central air conditioning terminal units and establishing a first neural network as a central air conditioning terminal unit metering mapping model according to the classification, classifying central air conditioning cooling rooms according to orientation and establishing a second neural network as a room cooling model for each orientation room, wherein the first neural network takes central air conditioning system operating parameters as input and the equivalent cooling capacity of the terminal unit per unit time in the room as output, the second neural network takes two scalars of the room temperature and outdoor air temperature during cooling, and a vector composed of the indoor temperature values in the six directions of front, back, left, right, up and down of the room as input, and takes the equivalent cooling capacity per unit time of the room as output, a reference air conditioning unit for cooling capacity comparison is selected based on the refrigeration capacity of the central air conditioning terminal unit, and the sensible cooling capacity of the reference air conditioning unit under different operating conditions is obtained according to standard tests and recorded as the operating characteristics; S2, collecting data samples and training the first neural network, controlling the terminal unit and the reference air conditioning unit to independently refrigerate in an alternating manner in stages, and dynamically maintaining the temperature and humidity of the room at preset target values in each stage, and taking the refrigeration capacity per unit time of the reference air conditioning unit under the same operating condition as the equivalent cooling capacity of the terminal unit; S3, for the second neural network, collecting data samples and training: when the operating condition is approximately stable, collect data samples of the second neural network under different input conditions, wherein the sample output is predicted by the trained first neural network, and the training is based on the data samples; S4, in online application, the central air conditioning is cost allocated in time periods: First, according to the cold consumption model, the correction coefficient is calculated under the current room condition where i is the current room number, the room number i = 1 ~ N is any integer, N is the total number of rooms, p i , p i0 is the mapping output of the second neural network when the input vector is , is the current condition of room i, the indoor temperature of the current room and the outdoor temperature are the same as , and the indoor temperature values of the six directions of the current room are the indoor temperature of the current room. The fee of the i-th room in the d-th time period is recalculated as: wherein, PF is the predicted cooling capacity of the end unit per unit time in the current room based on the first neural network i (t) the cumulative cooling capacity of the current period, Q jd Qk is the cooling capacity of room j in the current period, k j3 Cj is the corresponding coefficient of room j d is the total cost of the central air conditioning to be allocated in the current period.

2. A central air conditioning billing method, comprising the following steps: S1, initialization, classifying central air conditioning terminal units and establishing a first neural network as a central air conditioning terminal unit metering mapping model according to the classification, classifying central air conditioning cooling rooms according to orientation and establishing a second neural network as a room cooling model for each orientation room, wherein the first neural network takes central air conditioning system operating parameters as input and the equivalent cooling capacity of the terminal unit per unit time in the room as output, the second neural network takes two scalars of the room temperature and outdoor air temperature during cooling, and a vector composed of the indoor temperature values in the six directions of front, back, left, right, up and down of the room as input, and takes the equivalent cooling capacity per unit time of the room as output, a reference air conditioning unit for cooling capacity comparison is selected based on the refrigeration capacity of the central air conditioning terminal unit, and the sensible cooling capacity of the reference air conditioning unit under different operating conditions is obtained according to standard tests and recorded as the operating characteristics; S2, collecting data samples and training the first neural network: controlling the terminal unit and the reference air conditioning unit to independently refrigerate in an alternating manner in stages, and dynamically maintaining the temperature and humidity of the room at preset target values in each stage, and taking the refrigeration capacity per unit time of the reference air conditioning unit under the same operating condition as the equivalent cooling capacity of the terminal unit; S3, collecting data samples and training the second neural network: collecting data samples of the second neural network under different input conditions when the working condition is approximately stable, wherein the sample output is obtained by predicting the trained first neural network, and the training is based on the data samples; S4, when applied online, the central air conditioner is costed by time period: First, calculate the conversion factor based on the current room conditions. In the formula, i is the current room number, and s i s j The areas of rooms i and j are respectively, where i and j are any integers from 1 to N, and N is the total number of rooms. i p j Rooms i and j are respectively... When used as input vectors, each corresponds to the mapping output of the second neural network. The current operating status of room i. and calculate the comfort coefficient where p ic is the room i with As the input vector, its mapping output of the corresponding second neural network, working conditions and The difference between them is that the temperature of the room is a preset comfortable temperature. The fee of the i-th room in the d-th time period is recalculated as: Wherein, k j1 , k j2 is the corresponding coefficient of room j, C d is the total cost of the central air conditioning to be allocated in this period.

3. The central air conditioning billing method according to claim 1 or 2, characterized by, When the central air conditioner is a water-cooled central air conditioner, the end unit is an end air damper, and the first neural network takes the central air conditioner host refrigerated water supply flow, supply and return water temperature difference, current temperature and humidity of the room, and a vector composed of all end air damper opening state values as input; When the central air conditioner is a full-air central air conditioner, the end unit is an end air damper, and the first neural network takes the central air conditioner host air supply flow, supply and return air temperature difference, current temperature and humidity of the room, and a vector composed of all end air damper opening state values as input; When the central air conditioner is a variable refrigerant flow central air conditioner, the end unit is an indoor unit, and the first neural network takes the central air conditioner host power, indoor unit power, current temperature and humidity of the room, and a vector composed of all end unit fan opening state values as input.

4. The central air conditioning billing method according to claim 1 or 2, wherein The data sample collection process in step S2 is: Taking the room where the central air conditioner end unit is located as the thermal load, setting the target temperature according to the initial temperature of the room; controlling the central air conditioner end unit fan driver to operate to adjust the temperature of the room to the target temperature and obtain the current humidity as the target humidity, After the room remains at the target temperature for a period of time, the reference air conditioning unit, the central air conditioner end unit, and the reference air conditioning unit are used as the cold source to work for a certain period of time to collect samples, and the room temperature is kept at the target temperature when the two kinds of cold sources are working, Based on the working characteristics and working conditions of the reference air conditioning unit, the cooling power is calculated and divided by the average PWM duty cycle of the central air conditioner end unit working period, and the obtained value is taken as the equivalent refrigerating capacity per unit time when the central air conditioner end unit fan is in the current opening state value F.

5. The central air conditioning billing method according to claim 1 or 2, wherein, The step S2 specifically includes: S21, turn off the central air conditioning terminal unit of the room and start timing, control the reference air conditioning unit to maintain the room temperature at the target temperature from t=0 to t=T1, and control the working condition adjusting unit to keep the room humidity at the target humidity, calculate the refrigeration power q(t) based on the working characteristics and the working condition, and accumulate the refrigeration amount of the first period S22, the reference air conditioning unit is closed and the timer is reset, the fan on state value F of the central air conditioning terminal unit is set, the fan corresponding driver is controlled to work in PWM mode, the room temperature is maintained at the target temperature from t=0 to t=T2, and the equivalent time is calculated wherein Δ(t) is the PWM value, S23, again turn off the central air conditioner terminal unit and restart timing, control the reference air conditioning unit to maintain the room temperature at the target temperature from t=0 to t=Tl, and control the working condition adjusting unit to maintain the room humidity at the target humidity, and again calculate the third period of time its refrigerating capacity S24、calculating the equivalent refrigerating capacity per unit time of the fan under the current working condition:

6. The method of claim 1, wherein, current room period consumption of cooling energy when the room has multiple end units where F i is the set of all damper positions of the fans in the current room; The cost allocation of the central air conditioner also includes an area allocation part, and the cost of the i-th room in the d-th time period is replaced by: Wherein, C d1 and C d0 are the operating cost and the basic cost of the central air conditioner, respectively.

7. The central air conditioning billing method according to claim 1 or 2, wherein The data samples of the second neural network are automatically extracted during the operation of the central air conditioner: the thermal load working condition and cooling consumption parameters of the central air conditioner cooling room are periodically and continuously collected, and the periods in which the input quantities of the second neural network change within the preset fluctuation threshold are searched out, and the data of the selected periods are stored in the data sample set of the second neural network.

8. The method of claim 7, wherein, The preset fluctuation threshold includes a first fluctuation threshold and a second fluctuation threshold, and when the change of the second neural network input quantity in a period exceeds the first fluctuation threshold and is less than the second fluctuation threshold, the sample value of the input quantity in the period is taken as the time average value of the input quantity in the period.

9. The central air conditioning billing method according to claim 1 or 2, wherein Among the input quantities of the first neural network, the chilled water supply flow of the host and the supply and return water temperature difference are obtained by measuring points arranged adjacent to the central air conditioner host at the end of the chilled water supply and return water main pipeline; In step S2, the terminal unit and the reference air conditioning unit independently cool in three stages in an alternating manner, and the temperature difference between the temperature detection modules at multiple measuring points in the room is less than a temperature difference threshold value, the temperature detection modules can be arranged at the same height and located on two perpendicular diagonal lines respectively, and the temperature difference threshold value can be a value between 0.1℃ and 0.5℃, and the current temperature value of the room is the average value of the temperature values of multiple measuring points or the temperature value of the return air inlet; The humidity is sensed by a humidity detection module arranged in the middle of the room return air duct; The on state value of the central air conditioner terminal unit can be respectively taken as the normalized value of the fan power corresponding to the low, medium and high three gears of the fan speed, for example, taking the highest power value as 1, and the other two power values are converted in proportion.

10. The central air conditioning billing method according to claim 1 or 2, wherein The step S2 further comprises: Collecting room images, extracting orientation features and two mutually perpendicular diagonal lines through image processing, the orientation features including the distribution and length of the room structure edge lines, and the direction and distance of the cold source air outlet, return air inlet and uniform temperature module relative to the corners of the room; Taking the line connecting the cold air outlet position to the farthest place in the room that it can reach or the opposite position of the opposite side as one of the main diagonal lines, and taking a straight line perpendicular to it as the other main diagonal line; then, planning a spiral trajectory with the main diagonal line as the axis; In step S2, the terminal unit and the reference air conditioning unit independently cool in three stages in an alternating manner, and the temperature difference between the temperature detection modules at multiple measuring points in the room is less than a temperature difference threshold value, the temperature detection modules can be arranged at the same height and located on two perpendicular diagonal lines respectively, and the temperature difference threshold value can be a value between 0.1℃ and 0.5℃, and the current temperature value of the room is the average value of the temperature values of multiple measuring points or the temperature value of the return air inlet; 11. The central air conditioning billing method according to claim 1 or 2, wherein In step S1, the working characteristics of the reference air conditioning unit are obtained based on a room air enthalpy method test device, and the sensible heat of the reference air conditioning unit under different working conditions is tested, and the working condition parameters and the sensible heat are recorded as a working characteristic table or curve; In step S2, when collecting the first neural network sample, the cooling capacity of the reference air conditioning unit under the current working condition is calculated by querying and interpolating the table or curve based on the current working condition parameter value.

12. The central air conditioning billing method according to claim 1 or 2, wherein, The first neural network adopts a BP neural network, and the model is: The output of the jth node of the hidden layer is The output of the output layer is Wherein, x1-x4 are four scalar values of chilled water supply flow, supply and return water temperature difference, current temperature and humidity of the room, x5-xn are all central air conditioning terminal unit opening state values; f() function is sigmoid function, w ij and v j are connection weights from input layer to hidden layer and from hidden layer to output layer, respectively, θ j and θ are hidden layer and output layer thresholds, respectively, n and k are input layer and hidden layer node numbers, and gradient descent method is used for network training.

Images